Smart Quota Orchestrator
<div align="center">
<img src="./assets/hero-showcase.svg" alt="Smart Quota & Superpowers OS v13.0 Zenith Showcase" width="100%" />
<br />
<p>
<a href="https://sqo.elnemer-ai.com"><strong>🌐 Live Interactive Studio & Docs (sqo.elnemer-ai.com)</strong></a>
•
<a href="./README.ar.md"><strong>🇸🇦 الدليل العربي الشامل والمفصّل</strong></a>
•
<a href="./docs/PLAYBOOK_EN.md"><strong>📘 Real-World Playbook (EN)</strong></a>
•
<a href="./docs/PLAYBOOK_AR.md"><strong>📗 كتاب السيناريوهات العملية (عربي)</strong></a>
•
<a href="https://sqo.elnemer-ai.com/apex-skill.md"><strong>📄 1-File Standalone CLAUDE.md</strong></a>
</p>
<p>
<a href="https://github.com/ElNeMeR70/smart-quota-orchestrator"><img src="https://img.shields.io/badge/Release-v13.0.0_Zenith-4F46E5?style=flat-square" alt="Release v13.0.0 Zenith" /></a>
<a href="./mcp_server.py"><img src="https://img.shields.io/badge/MCP_Server-53_Native_Tools-059669?style=flat-square" alt="53 Native MCP Tools" /></a>
<a href="./sqo.py"><img src="https://img.shields.io/badge/Master_CLI-67_Commands-0284C7?style=flat-square" alt="67 CLI Commands" /></a>
<a href="./tests/test_scripts.py"><img src="https://img.shields.io/badge/Automated_Tests-66%2F66_Passing-10B981?style=flat-square" alt="66/66 Tests Passing" /></a>
<a href="./package.json"><img src="https://img.shields.io/badge/Dependencies-Zero_(100%25_Python_Stdlib)-D97706?style=flat-square" alt="Zero Dependencies" /></a>
<a href="./LICENSE"><img src="https://img.shields.io/badge/License-MIT-7C3AED?style=flat-square" alt="MIT License" /></a>
</p>
</div>
---
## ⚡ 30-Second Quickstart (Zero `pip` Packages Required!)
Pick **any** workflow below — works out-of-the-box with **Google Antigravity**, **Claude Code**, **Cursor**, **VS Code Copilot**, **Gemini CLI**, and **Windsurf**:
```bash
# Option 1 — Universal 1-Command Project Bootstrap (Auto-detects Flutter / Next.js / Laravel / Python / Go / Rust)
npx -y smart-quota-orchestrator init --workspace .
# Option 2 — 1-Line Cloud Installer (Windows PowerShell)
irm https://sqo.elnemer-ai.com/install.ps1 | iex
# Option 3 — 1-Line Cloud Installer (macOS / Linux)
curl -fsSL https://sqo.elnemer-ai.com/install.sh | bash
# Option 4 — Just want a drop-in 1-File Constitution (Karpathy + FinOps + Quota Rules)?
curl -fsSL https://sqo.elnemer-ai.com/apex-skill.md -o CLAUDE.md
```
Once installed, run **one single command** on any bug or feature (in English or Arabic):
```bash
# Compiles Repo-DNA + BM25 line coordinates + Hidden Git Co-Change Twins + Pinned SDK Idioms in <45ms!
npx smart-quota-orchestrator autopilot "Fix duplicate Stripe webhook charge & slow orders dashboard" --workspace .
# Before committing or opening a PR: run all 8 deterministic security, FinOps, RTL & performance gates in <65ms!
npx smart-quota-orchestrator pr-shield --workspace .
```
---
## 🔥 Why Do AI Coding Quotas Evaporate in 35 Minutes? (And How `v13.0` Fixes It)
Modern AI models (**Claude Opus / Sonnet 4.6**, **Gemini 4 Carbon / Hydrogen**, **GPT**) are extraordinary engineers, yet **85%–95% of your 5-hour token quota and API bill is burned on invisible runtime waste** — not the 30 lines of code you actually ship. Worse, unchecked AI diffs frequently introduce **silent `$5,000` cloud billing loops**, **SQL injection**, **TOCTOU wallet race conditions**, and **broken Arabic RTL layouts**.
<p align="center">
<img src="./assets/hero-banner.jpg" alt="Smart Quota & Superpowers OS Aurora Daylight Visual Core" width="100%" />
</p>
### 📊 Head-to-Head Comparison Across the AI Developer Ecosystem
| Capability | Unprotected AI IDE | Static `.cursorrules` | Repo Dumpers (`Repomix`) | Single-Purpose MCPs | **⚡ Smart Quota OS `v13.0`** |
| :--- | :---: | :---: | :---: | :---: | :---: |
| **Blocks `>50KB` Lockfiles & `.map` Pre-Billing** | ❌ Burns `120k` tok | ⚠️ Ignored 40% of time | ⚠️ Dumps huge text | ❌ No | **✅ Deterministic `<15ms` Hook** |
| **Prevents Re-Reading Unchanged Files 3–5×** | ❌ Re-reads blindly | ❌ Cannot track state | ❌ Static dump | ❌ No | **✅ Stateful `mtime_ns` Lock** |
| **Symbol-Level AST Edits (`-85%` Output Tokens)** | ❌ Rewrites full file | ❌ Text prompt only | ❌ Read-only | ✅ (`Serena` + LSP deps) | **✅ `sqo_edit_symbol` (`0` deps)** |
| **Single-Shot `<420-Token` Repo Architecture DNA** | ❌ `25,000+` tok walk | ❌ Manual writing | ❌ `40k–200k` tok | ❌ No | **✅ `sqo_repo_dna` (`-99.2%`)** |
| **Cloud & AI FinOps `$5,000` Bill-Shock Guard** | ❌ Ships infinite loops | ❌ Passive text | ❌ No | ❌ No | **✅ `sqo_finops_guard`** |
| **4-Seat Adversarial Council (TOCTOU + SRE)** | ❌ Single-agent bias | ❌ No | ❌ No | ❌ No | **✅ `sqo_council_review` (`0` tok)** |
| **OWASP SAST + N+1 DB Profiler + Arabic RTL** | ❌ Ships subtle bugs | ❌ Passive text | ❌ No | ⚠️ Separate heavy tools | **✅ 8-Gate `sqo_pr_review`** |
| **Git Evolutionary Hidden Co-Change Twins** | ❌ Misses non-imports | ❌ No | ❌ No | ❌ No | **✅ `sqo_git_forensics`** |
| **External Dependencies / Background RAM** | — | `0` | Node/Python deps | Heavy LSP / Vector DBs | **✅ `100%` Stdlib • `0.0%` Idle CPU** |
---
## 🏗️ Deep Dive: Inside the 6-Layer Zenith Architecture
<p align="center">
<img src="./assets/architecture-v5.svg" alt="Smart Quota & Superpowers OS v13.0 Architecture" width="100%" />
</p>
### Layer 1 — The 3-Stage `<15ms` Stateful Runtime Hook Firewall (`-96.8%` Input Waste)
Passive Markdown rules *ask* an LLM not to waste tokens; under long-context pressure, LLMs forget. [`hooks/token_shield_hook.py`](./hooks/token_shield_hook.py) enforces strict token discipline **at the OS tool-execution boundary before tokens hit your bill**:
<p align="center">
<img src="./assets/shield-lifecycle.svg" alt="3-Stage Stateful Hook Firewall" width="100%" />
</p>
1. **`SessionStart` (Zero Cold-Start Tax + 100% KV-Cache Protection)**:
Injects your top 8 project architecture memories (`.sqo/memory.db`) + federated cross-repo wisdom (`~/.sqo/hive_mind.db`) **once** at Turn 0. Unlike dynamic per-turn hooks that inject timestamps and destroy your **90% LLM Prompt KV-Cache discount**, SQO keeps your system prefix **100% byte-identical** across turns.
2. **`PreToolUse` (4-in-1 Pre-Billing Interceptor)**:
- **Token-Bomb Blocker**: Denies unbounded reads or `cat` dumps on `package-lock.json`, `pubspec.lock`, `pnpm-lock.yaml`, `.min.js`, and `.js.map`.
- **Stateful Duplicate-Read Deduplicator**: Hashes `(file_path, mtime_ns, line_range)`. If the agent tries to re-read an unchanged file already in context, SQO halts the duplicate read instantly and unlocks it automatically the millisecond `edit_file` modifies it!
- **Giant-Monolith Auto-Clamper**: Transparently rewrites blind full-file reads on `>1,200`-line source files into bounded `StartLine: 1, EndLine: 280` reads.
- **Blind Retry-Loop Breaker**: Halts the 3rd identical failing terminal command if zero files were edited in between.
3. **`PostToolUse` (Silent Auto-Memory Graph & `$USD` ROI Ledger)**:
Silently records file co-edits and `test failed -> edit -> test passed` bugfix resolutions into `.sqo/memory_graph.db`, while logging every single saved token and `$USD` equivalent into `.sqo/roi_ledger.db`!
<p align="center">
<img src="./assets/benchmark-comparison.svg" alt="Empirical Before-vs-After Token Benchmark" width="100%" />
</p>
---
### Layer 2 — `sqo autopilot`, `<420-Token` Repo-DNA & Surgical AST Editing (`-85%` Output Tokens)
Output tokens cost **3× to 5× more** than input tokens on frontier models (`$15–$75 / 1M tokens`), yet most agents rewrite an entire 500-line file just to change one 18-line method!
- **`sqo autopilot "<goal>"` ([`autopilot_engine.py`](./skills/smart-quota-orchestrator/scripts/autopilot_engine.py))**: Runs `Repo-DNA` + `Prompt X-Ray` (BM25 symbol pinning in English & Arabic) + `Git Forensics` + `SDK Radar` in `<45ms` and hands the agent an exact surgical mission brief.
- **`sqo_edit_symbol` ([`edit_symbol.py`](./skills/apex-spec-surgeon/scripts/edit_symbol.py))**: Locates `ClassName.method_name` via AST/brace boundaries, takes an automatic Time-Machine backup, splices **only** the mutated symbol (**saving ~85% of Output Tokens!**), verifies syntax in `<8ms` (`ast.parse` / brace & string balance), and **auto-rolls back** if a syntax error is detected!
- **`sqo rules-gen` ([`compile_repo_rules.py`](./skills/smart-quota-orchestrator/scripts/compile_repo_rules.py))**: Compiles a bespoke, repo-aware `CLAUDE.md` / `AGENTS.md` / `.cursorrules` tailored to your exact `package.json` / `pubspec.yaml` scripts, core modules, and SDK versions in `<25ms` (**`0` LLM tokens burned**, unlike `claude /init` which burns `30,000+` tokens!).
<p align="center">
<img src="./assets/terminal-roi-preview.svg" alt="Live Terminal Preview of sqo autopilot, sqo pr-shield and sqo roi" width="100%" />
</p>
---
### Layer 3 — Cloud & AI FinOps Bill-Shock Guard (`sqo_finops_guard`)
AI coding agents love writing code that "works on localhost" while planting **catastrophic cloud billing time-bombs**. [`finops_guard.py`](./skills/apex-spec-surgeon/scripts/finops_guard.py) catches them statically before you deploy:
```tsx
// ❌ BLOCKED BY sqo_finops_guard [FINOPS-INF-USEEFFECT • Risk: $2,500/day]:
// Missing `[]` dependency array causes setOrders() -> re-render -> new Firestore listener -> infinite $$$ loop!
useEffect(() => {
onSnapshot(collection(db, "orders"), (snap) => setOrders(snap.docs));
});
// ❌ BLOCKED BY sqo_finops_guard [FINOPS-RECURSIVE-TRIGGER • Risk: $10,000/hr]:
// Cloud Function triggered on `invoices` writes back to `invoices` with no idempotency guard!
export const onInvoice = onDocumentWritten("invoices/{id}", async (e) => {
await db.collection("invoices").add({ syncedAt: Date.now() });
});
```
Also flags **paid LLM API calls inside `for`/`while` loops**, **LLM calls missing `max_tokens` / `max_output_tokens` caps**, and **public AI HTTP endpoints missing auth/rate-limiting**!
---
### Layer 4 — 8-Gate Pre-PR Autonomous Code Doctor (`sqo pr-shield` / `sqo_pr_review`)
Before committing or merging any AI-generated code, [`pr_shield.py`](./skills/apex-spec-surgeon/scripts/pr_shield.py) runs **8 deterministic engineering audits in `<65ms`** and generates a complete GitHub Pull Request Review Table:
1. **🛡️ OWASP Security SAST (`sqo_security_audit`)**: Catches SQL Injection (`CWE-89`), Command Injection (`CWE-78`), XSS (`CWE-79`), `eval()` RCE (`CWE-95`), and Weak `MD5/SHA1` password hashing across Python, JS/TS, PHP & Dart.
2. **💸 Cloud & AI FinOps Guard (`sqo_finops_guard`)**: Zero infinite `useEffect` read loops, recursive triggers, or uncapped LLM loops.
3. **🎭 4-Seat Adversarial Council (`sqo_council_review`)**:
- **🔴 Exploit Hunter**: Catches **TOCTOU Double-Spend Race Conditions** (`if user.balance >= amt: user.balance -= amt` in memory instead of atomic SQL `UPDATE wallets SET balance = balance - ? WHERE id = ? AND balance >= ?`).
- **🟠 3 AM Principal SRE**: Flags outbound `requests.post(...)` / `httpx` calls missing `timeout=` (which freeze production workers indefinitely) and silent `except: pass` / `catch {}` blocks.
- **🔵 UX Critic & 🟢 Karpathy Minimalist**: Catches leftover debug `console.log`/`debugger` noise.
4. **⚡ N+1 & UI-Jank Profiler (`sqo_perf_audit`)**: Catches DB/HTTP queries inside loops, sequential `await` waterfalls, undisposed Flutter `TextEditingController`s, and `jsonDecode` inside `Widget build()`.
5. **🎨 Bilingual Arabic RTL & WCAG Auditor (`sqo_ui_audit`)**: Catches hardcoded `EdgeInsets.only(left:)` / `margin-left:` that break Arabic RTL layouts, unexpanded `Text` inside Flutter `Row`s, and `<img>` tags missing `alt`.
6. **⚖️ Karpathy Gate & Live SDK Radar (`sqo_verify_work` + `sqo_sdk_radar`)**: Blocks `TODO` laziness stubs, silent function deletions, and deprecated 2023 SDK hallucinations (`WillPopScope`, `.withOpacity()`, legacy `google.generativeai`, Next.js 15 sync `params`).
7. **📦 Anti-Hallucination Dependency & `.env` Parity (`sqo_verify_deps` + `sqo_drift_guard`)**: Catches undeclared phantom imports and missing `.env.example` keys before boot crashes.
8. **🔐 10-Rule Secret DLP (`sqo_secret_scrub`)**: Blocks leaked OpenAI/Anthropic/Gemini keys, GitHub PATs, JWTs, Private Keys, and DB URIs.
---
### Layer 5 — Git Evolutionary Forensics, Live Mermaid Diagrams & Hive-Mind Memory
- **🕵️ Git Evolutionary Forensics (`sqo forensics` / `sqo_git_forensics`)**: Static AST graphs can only see explicit `import` statements. `git_forensics.py` mines `git log` commit co-occurrence matrices to uncover **Hidden Co-Change Twins** — e.g., `backend/stripe_webhook.py` and `mobile/lib/models/plan_tier.dart` that share `0` imports yet co-change in `82%` of commits — plus **Historical Bug-Magnet Hotzones**!
- **🕸️ Zero-Token Live Mermaid Generator (`sqo arch-mermaid` / `sqo_arch_diagram`)**: Auto-generates 3 copy-pasteable Mermaid diagrams (`flowchart LR` Core Gravity Modules, `flowchart TD` HTTP/UI Route Map, and `erDiagram` Database Schema) in `<25ms`!
- **🧠 Cross-Project Federated Hive-Mind (`sqo hive` / `sqo_hive_mind`)**: Promote hard-won stack lessons once into `~/.sqo/hive_mind.db` so **every new repository** on your machine inherits your engineering wisdom automatically!
---
## 🛠️ Step-by-Step Integration Guide by AI IDE
<details open>
<summary><strong>1. Cursor IDE (<code>.cursor/mcp.json</code> + <code>.cursorignore</code> + Bespoke Rules)</strong></summary>
```bash
npx -y smart-quota-orchestrator sync-ide --workspace . --ides cursor
npx -y smart-quota-orchestrator rules-gen --workspace . --write .cursorrules
```
Automatically registers `sqo-engine` (all 53 MCP tools) in `.cursor/mcp.json`, generates stack-aware `.cursorignore` rules blocking lockfiles/build artifacts, and compiles a custom `.cursorrules` tailored to your repo!
</details>
<details>
<summary><strong>2. Claude Code CLI (<code>.mcp.json</code> + <code>CLAUDE.md</code> + Pre/Post Hooks)</strong></summary>
```bash
npx -y smart-quota-orchestrator sync-ide --workspace . --ides claude
npx -y smart-quota-orchestrator rules-gen --workspace . --write CLAUDE.md
```
Wires `.mcp.json`, `.claudeignore`, and a repo-specific `CLAUDE.md` constitution in under 1 second.
</details>
<details>
<summary><strong>3. Google Antigravity (Full Native Plugin: 53 MCP Tools + 3-Stage Hooks + 7 Skills + 3 Subagents)</strong></summary>
```powershell
# Windows PowerShell
irm https://sqo.elnemer-ai.com/install.ps1 | iex
# macOS / Linux
curl -fsSL https://sqo.elnemer-ai.com/install.sh | bash
```
Installs the complete plugin (`plugin.json`), activates `SessionStart` / `PreToolUse` / `PostToolUse` stateful hooks, and unlocks specialist subagents (`codebase-scout`, `quota-planner`, `token-auditor`).
</details>
<details>
<summary><strong>4. VS Code Copilot & Windsurf + GitHub Actions CI Shield</strong></summary>
```bash
npx -y smart-quota-orchestrator sync-ide --workspace . --ides vscode,windsurf
npx -y smart-quota-orchestrator git-guard --workspace .
npx -y smart-quota-orchestrator ci-shield --workspace .
```
Configures `.vscode/mcp.json`, `.codeiumignore`, local `<40ms` Git `pre-commit` secret/tax protection, and automated PR Shield checks!
</details>
---
## 🔌 Complete Catalog of All 53 Zero-Dependency Native MCP Tools
| Category | Native MCP Tools (`sqo-engine`) |
| :--- | :--- |
| **🚀 Autonomous Orchestration & Specs (8)** | `sqo_autopilot` • `sqo_spec_engine` • `sqo_compile_rules` • `sqo_init_preset` • `sqo_swarm_plan` • `sqo_dependency_slices` • `sqo_quota_pools` • `sqo_context_health` |
| **🧬 Ultra-Lean Code Recon & Diagrams (10)** | `sqo_repo_dna` • `sqo_code_search` • `sqo_smart_outline` • `sqo_find_callers` • `sqo_code_hotspots` • `sqo_api_routes` • `sqo_arch_diagram` • `sqo_git_forensics` • `sqo_blast_radius` • `sqo_pinpoint_error` |
| **🗡️ Surgical Editing & Zero-Risk Sandboxes (5)** | `sqo_edit_symbol` (`-85%` Out Tok) • `sqo_checkpoint` (Time-Machine) • `sqo_worktree_sandbox` • `sqo_gen_tests` (`0`-Tok Synth) • `sqo_shadow_check` |
| **🛡️ 8-Gate Security, FinOps & Quality (11)** | `sqo_pr_review` • `sqo_finops_guard` • `sqo_security_audit` • `sqo_council_review` • `sqo_perf_audit` • `sqo_ui_audit` (RTL/a11y) • `sqo_verify_work` (Karpathy) • `sqo_sdk_radar` • `sqo_verify_deps` • `sqo_drift_guard` • `sqo_dead_code` |
| **🗜️ Context, Schema & Media Compaction (10)** | `sqo_prompt_xray` • `sqo_context_pack` • `sqo_schema_distill` (6-Stack DB) • `sqo_api_contracts` (OpenAPI 3.1) • `sqo_budgeted_diff` • `sqo_compress_prompt` • `sqo_media_budget` • `sqo_i18n_audit` • `sqo_cache_audit` • `sqo_secret_scrub` |
| **🧠 Memory, ROI Telemetry & Cloud Hub (9)** | `sqo_memory_vault` • `sqo_memory_graph` • `sqo_hive_mind` • `sqo_memo_cache` (`2ms` SHA-256) • `sqo_roi_ledger` • `sqo_benchmark` • `sqo_cloud_hub` • `sqo_repo_badge` • `sqo_ci_shield` |
---
## ❓ Frequently Asked Questions (FAQ)
- **Does SQO send my source code to any external server or LLM API?**
**100% Local & Offline by default.** All 53 MCP tools, SQLite databases (`.sqo/`), BM25 search, SAST, FinOps, and hook shields run locally on your machine in pure Python 3.8+ standard library (`0` external `pip` packages). Cloud features (`sqo hub pull` or `sqo share --publish`) are strictly opt-in.
- **How fast are the runtime hooks and tools?**
`PreToolUse` hook checks run in **`4ms–14ms`**, incremental `mtime_ns` SQLite graph warm queries run in **`<10ms`**, and a full 8-Gate `sqo pr-shield` workspace audit finishes in **`<65ms`**.
- **Does it break Anthropic / OpenAI / Gemini Prompt KV-Caching?**
Never — SQO is specifically engineered to **protect your 90% KV-Cache discount** (`sqo cache-audit` verifies `100/100` static prefix stability).
---
<div dir="rtl">
## 🇸🇦 الملخص العربي السريع
تُعد منظومة **Smart Quota & Superpowers OS (`v13.0 Zenith`)** أول نظام تشغيل هندسي متكامل ومجاني مفتوح المصدر (`Zero Dependencies`) يجمع بين **مضاعفة كوتا وكلاء الذكاء الاصطناعي من 10 إلى 25 ضعفاً** وبين **حماية مشاريعك في الإنتاج من الثغرات والفواتير الكارثية**:
- 🚀 **أمر القيادة الذاتية الموحد (`sqo autopilot`)**: يفحص البصمة الجينية للمشروع (`Repo-DNA`)، ويحدد أسطر الدوال المطلوبة بدقة (`BM25`)، وينبّهك للتوائم الخفية في تاريخ `Git`، ويحقن قواعد الإصدارات الحديثة في **أقل من `45ms` بصفر توكن**!
- 💸 **حارس فواتير السحابة والـ AI (`sqo finops`)**: يمنع كوارث الفواتير اللانهائية قبل الرفع (مثل `useEffect` مع قراءات `Firestore` بدون `[]`، والدوال السحابية العودية، واستدعاءات الـ LLM داخل الحلقات).
- 🛡️ **بوابة فحص ما قبل النشر الثمانية (`sqo pr-shield`)**: تفحص 8 محاور هندسية دفعة واحدة (ثغرات `SQLi/XSS`، ثغرات السحب المزدوج للرصيد `TOCTOU`، أخطاء الاتجاه العربي `RTL`، بطء استعلامات `N+1`، المكتبات المهلوسة، وتسريب المفاتيح السرية).
- 🗡️ **التعديل الجراحي للدوال (`sqo_edit_symbol`) + مولّد القواعد الذكي (`sqo rules-gen`)**: يوفر **85% من توكنات الكتابة الباهظة** ويولّد ملف `CLAUDE.md` / `.cursorrules` مخصص لمشروعك في ثانية واحدة!
📖 **اقرأ الشرح العربي التفصيلي الكامل والأمثلة العملية خطوة بخطوة في:**
- **[🇸🇦 الموسوعة العربية الشاملة (`README.ar.md`)](./README.ar.md)**
- **[📗 كتاب الحوارات والسيناريوهات الواقعية (`docs/PLAYBOOK_AR.md`)](./docs/PLAYBOOK_AR.md)**
- **[🌐 المنصة التفاعلية الحية (بالعربية والإنجليزية): `https://sqo.elnemer-ai.com`](https://sqo.elnemer-ai.com)**
</div>
TDQS
Scored across 53 tools
The set has many overlapping clusters: sqo_pr_review, sqo_security_audit, sqo_council_review, and sqo_verify_work all perform diff/code audits; sqo_memory_vault, sqo_hive_mind, sqo_memory_graph, and sqo_memo_cache all store/retrieve memory; sqo_context_pack, sqo_repo_dna, sqo_compress_prompt, and sqo_prompt_xray all compress or package context. Some tools are genuinely distinct, but the boundaries between several categories are unclear enough that an agent could easily choose the wrong audit, memory, or compression tool.
Every tool uses the same sqo_ snake_case prefix, which makes the set easy to recognize and scan. However, the suffix pattern mixes noun phrases (sqo_dead_code, sqo_memory_vault, sqo_blast_radius) with verb-led names (sqo_compile_rules, sqo_verify_deps, sqo_edit_symbol), so it is mostly consistent rather than perfectly verb_noun throughout.
With 53 tools, the surface is far beyond a well-scoped MCP set and is likely to overwhelm tool selection. Many tools are near-duplicates or subsumed by orchestrator tools like sqo_pr_review and sqo_autopilot, so the count is an extreme mismatch for the server's purpose.
The surface broadly covers repo analysis, context compression, auditing, memory, planning, quota tracking, test generation, and symbol editing, so most workflows in the stated domain have a tool. There are minor gaps around general whole-file read/write/list operations and direct build/test execution, but agents can work around these with the symbol-editing and analysis tools.